persists, that is, there were not sufficient records to be sure of including the smaller storms. For this reason, therefore, detail consideration was limited to the two greatest storms of that period: that of July 27-31, 1887, central over Georgia, and that of May 31-June 1, 1889, central over Pennsylvania.

On July 1, 1891, the Weather Bureau of the United States Department of Agriculture took over the climatological work which for twenty years had been conducted by the Signal Service of the War Department. A large number of additional observing stations were soon established, especially in those parts of the country where but few had existed prior to that time. We can feel sure, for the 25-year period 1892-1916, that not only have we records of all the storms that have occurred which come within the selected limits, but also that the data is sufficient to warrant a study of comparative sizes by means of their time-area-depth relations, and a study of their average frequency and seasonal distribution.

It is to this 25-year period that the greatest amount of study has been given. In the determination of storm frequency, and seasonal and geographical occurrence in the next chapter, no attempt is made. to use the less complete data for the 50 years prior to 1892; and in discussing in detail in a later chapter the time, area, and depth storm factors, consideration is limited, for this 50-year period, to only the three great storms just mentioned. We do not believe we are justified, however, in ignoring altogether the data for those early years It has very decided value in supplementing and corroborating the storm experience of the past 25 years, and will be further discussed in a later chapter in connection with the time-area-depth relations of 30 of the most important storms of the past 25 years.

SELECTING THE STORMS

In selecting the more important from the large number of storms which were recorded during the 25-year period, 1892-1916, it was necessary to fix maximum limits of area and duration which were to be treated as being comprised in a single storm. Obviously, it was also necessary to fix minimum limits of depth and area in order to exclude the numerous storms of such small area and depth as to be of little or no consequence in an investigation of this nature. The fixing of these limits was done somewhat arbitrarily, the principal object being so to choose them as to include all storms which could possibly be of interest. The criterion adopted was that each storm selected should have not less than five station records having a 3-day precipitation equaling or exceeding 6 inches.

Of such storms, 160 were found which occurred during the years 1892-1916. A list of these, divided into two groups, northern and southern, is given in tables 4 and 5, giving for each storm the identification number, date, geographical location of principal center, as well as the highest, fifth highest, tenth highest, and twentieth highest 3-day precipitation records. The storms marked with asterisks are the largest and most important and of these a detail study as to their time-area-depth relations is made in a subsequent chapter.

The actual location of all the storms of which there are records, and which come within the limits just described, required a careful and extended search of several sources. Most of them were located by consulting the monthly rainfall chart and the notes on floods in the Monthly Weather Review.* On the chart appear isohyetals showing the precipitation over the entire United States for the current month. Notes in the text generally supplement the chart, and explain in some detail the nature and extent of unusual storms. Since these charts show the amount of rainfall for an entire month, it occasionally happens that although there is a large rainfall over an extensive area it is so distributed throughout the month that the maximum five consecutive days appear as a storm of little or no consequence. In such cases as this, the notes in the text on storms and floods sometimes indicate the character of the precipitation, but it was generally necessary to refer to the daily rainfall records of stations in the storm area to determine whether the precipitation shown by the chart was distributed over an extended period of time, or fell principally as a storm sufficiently intense to come within the limits chosen. Months for which the rainfall chart indicated that no storm of the defined intensity could have occurred were passed by without further search.

DETERMINING THE RELATIVE SIZES OF THE STORMS It was early observed that the largest storms of the north and east never approach in depth and intensity the maximum Gulf coast and southern Atlantic seaboard storms. To establish an equitable basis for comparison of size all the storms were divided into two groups, the northern group and the southern group. The line of division was somewhat arbitrarily chosen along the north boundaries of North Carolina, Tennessee, Arkansas, and Oklahoma. Of the 160 storms, 47 are in the northern, and 113 in the southern group. This

*The Signal Service of the War Department first published the Monthly Weather Review in January 1873 and continued to issue it monthly until 1892, when the Weather Bureau of the U. S. Department of Agriculture was formed to take over the meteorological work of the Signal Service. Since 1892 the Weather Review has been the principal official periodical of the Weather Bureau.

Table 4.-Chronological List of 49 Great Northern Storms. The greatest, fifth, tenth, and twentieth highest rainfall records in inches for the maximum period of 3 days in each storm are given. The asterisks (*) denote important storms of which the time-area-depth relations were studied in detail by means of maps and curves.

Table 5.-Chronological List of 114 Great Southern Storms

The greatest, fifth, tenth, and twentieth highest rainfall records in inches for the maximum period of 3 days in each storm are given. The asterisks (*) denote important storms of which the time-area-depth relations were studied in detail by means of maps and curves.